KR20130014521A - Decoding apparatus, decoding method, encoding apparatus, encoding method, and program - Google Patents

Abstract

The present invention relates to a decoding apparatus and decoding method, an encoding apparatus and an encoding method, and a program capable of reducing the delay time due to the band extension during decoding and suppressing an increase in resources on the decoding side. The high pass component generator 73 generates a pseudo high pass spectrum using the low pass spectrum SP-L and the high pass envelope ENV-H. The phase random unit 74 randomizes the phase of the pseudo high frequency spectrum based on the random flag RND. The inverse MDCT unit 75 denormalizes the low pass spectrum SP-L using the low pass envelope ENV-L, and the pseudo high pass spectrum and the denormalized low pass spectrum SP supplied from the phase random portion 74. -L) is synthesize | combined, and the result of synthesis is made into the spectrum of the full band. The present invention can be applied to, for example, a decoding device that performs band extension decoding.

Description

Decoding device and decoding method, encoding device and encoding method, and program {DECODING APPARATUS, DECODING METHOD, ENCODING APPARATUS, ENCODING METHOD, AND PROGRAM}

The present invention relates to a decoding apparatus, a decoding method, an encoding apparatus and an encoding method, and a program. In particular, the present invention relates to reducing a delay time caused by bandwidth expansion during decoding and suppressing an increase in resources on a decoding side. The present invention relates to a decoding apparatus and decoding method, an encoding apparatus and an encoding method, and a program.

As a method of encoding a speech signal, transform coding methods such as Moving Picture Experts Group Audio Layer-3 (MP3), Advanced Audio Coding (AAC), and Adaptive Transform Acoustic Coding (ATRAC) are generally well known.

In such an encoding method, it is considered to improve the coding efficiency by not including the high frequency spectrum having a large amount of information in the encoding result, but including only the envelope of the high frequency spectrum. In this case, during decoding, the high frequency spectrum is generated by copying the low frequency spectrum by parallel movement or return. Then, only the generated high frequency envelope is approached to the original high frequency spectrum envelope included in the encoding result, so that the audio quality can be improved. This decoding technique is called a band extension technique and is already generally recognized.

1 is a block diagram showing an example of a configuration of an encoding device in which only an envelope is included in an encoding result for a high frequency spectrum.

The encoding device 10 of FIG. 1 includes a Modified Discrete Cosine Transform (MDCT) unit 11, a quantization unit 12, and a multiplexer 13. In addition, the encoding apparatus 10 is similar to the already known transformation encoding apparatus, except that the high frequency spectrum SP-H is not included in the encoding result. In addition, in order to simplify description of drawing, the quantization part 12 shall perform not only quantization but also extraction and normalization of a quantization object.

Specifically, the MDCT unit 11 of the encoding apparatus 10 performs MDCT on a PCM (Pulse Code Modulation) signal, which is a time domain signal of speech input to the encoding apparatus 10, to perform a spectrum (frequency spectrum signal). SP). The MDCT unit 11 supplies the generated spectrum SP to the quantization unit 12.

The quantization part 12 extracts an envelope from the high frequency spectrum SP-H which is the high frequency component of the spectrum SP supplied from the MDCT part 11, and the low frequency spectrum SP-L which is the low frequency component, respectively. The quantization unit 12 quantizes the high pass envelope ENV-H, which is the envelope of the extracted high pass spectrum SP-H, and the low pass envelope ENV-L, which is the envelope of the low pass spectrum SP-L. The quantization unit 12 supplies the quantized high pass envelope ENV-H and low pass envelope ENV-L to the multiplexer 13. In addition, in this specification, in order to simplify description, the names (SP-L, SP-H, etc.) of signals before and after quantization and encoding are made the same.

In addition, the quantization unit 12 normalizes the low pass spectrum SP-L using the low pass envelope ENV-L, quantizes the normalized low pass spectrum SP-L, and the low pass obtained as a result. The spectrum SP-L is supplied to the multiplexer 13.

As described above, the quantization unit 12 includes a spectrum normalized with the envelope in the encoding result for the low pass component of the spectrum SP, but includes only the envelope in the encoding result for the high pass component. This improves the coding efficiency.

The multiplexer 13 multiplexes the low pass envelope (ENV-L), low pass spectrum (SP-L) and high pass envelope (ENV-H) supplied from the quantization unit 12, and outputs the resulting bit stream. . This bit stream is recorded in a recording medium (not shown) or transmitted to the decoding device.

FIG. 2 is a flowchart for describing an encoding process by the encoding device 10 of FIG. 1. This encoding process is started, for example, when the PCM signal of audio | voice is input to the encoding apparatus 10. FIG.

In step S11 of FIG. 2, the MDCT unit 11 performs MDCT on a PCM signal that is a time domain signal of speech input to the encoding apparatus 10 to generate a spectrum SP that is a frequency domain signal. The MDCT unit 11 supplies the generated spectrum SP to the quantization unit 12.

In step S12, the quantization part 12 is enveloped from the high frequency spectrum SP-H which is the high frequency component of the spectrum SP supplied from the MDCT part 11, and the low frequency spectrum SP-L which is the low frequency component, respectively. Extract

In step S13, the quantization unit 12 normalizes the low pass spectrum SP-L using the low pass envelope ENV-L.

In step S14, the quantization unit 12 quantizes the extracted high pass envelope ENV-H, low pass envelope ENV-L, and normalized low pass spectrum SP-L. The quantization unit 12 supplies the quantized high pass envelope ENV-H, the low pass envelope ENV-L, and the normalized low pass spectrum SP-L to the multiplexer 13.

In step S15, the multiplexing unit 13 multiplexes the low pass envelope (ENV-L), low pass spectrum (SP-L), and high pass envelope (ENV-H) supplied from the quantization unit 12, and is obtained as a result. Output a bit stream. Then, the process ends.

FIG. 3 is a block diagram illustrating an example of a configuration of a decoding device that decodes a bit stream encoded by the encoding device 10 of FIG. 1.

The decoding apparatus 30 of FIG. 3 is comprised from the decomposition part 31, the inverse quantization part 32, the inverse MDCT part 33, and the band expansion part 34. As shown in FIG.

The decomposition section 31, the inverse quantization section 32, and the inverse MDCT section 33 of the decoding device 30 restore only the low-pass component of the PCM signal in the same manner as in the ordinary conversion decoding device.

Specifically, the decomposition unit 31 obtains the bit stream encoded by the encoding device 10, and converts it into a low pass envelope (ENV-L), a low pass spectrum (SP-L), and a high pass envelope (ENV-H). It decomposes | disassembles and supplies it to the dequantization part 32.

The inverse quantization unit 32 performs inverse quantization on each of the low pass envelope ENV-L, the low pass spectrum SP-L, and the high pass envelope ENV-H supplied by the decomposition unit 31. The inverse quantization unit 32 supplies the inversely quantized low pass envelope (ENV-L) and low pass spectrum (SP-L) to the inverse MDCT unit 33, and supplies the high pass envelope (ENV-H) to the band extension unit. It is supplied to 34.

The inverse MDCT unit 33 denormalizes the low frequency spectrum SP-L using the low frequency envelope ENV-L supplied from the inverse quantization unit 32. In addition, the inverse MDCT unit 33 performs inverse MDCT on the low frequency spectrum (SP-L), which is a denormalized frequency domain signal, to obtain a PCM signal which is a time domain signal. In addition, this PCM signal is a PCM signal without a high frequency component, and is a PCM signal which is audible indefinite sound. The inverse MDCT unit 33 supplies this PCM signal to the band extension unit 34.

The band extension section 34 is composed of a band division filter 41, a high pass component generator 42, and a band synthesis filter 43. The band extension section 34 expands the frequency band of the PCM signal without the high band component obtained by the inverse MDCT section 33, thereby performing band extension processing for improving the sound quality of the PCM signal.

Specifically, the band dividing filter 41 of the band extension section 34 divides the PCM signal supplied from the inverse MDCT section 33 into a high band component and a low band component. Since the PCM signal does not have a high frequency component, the band dividing filter 41 discards the high frequency component of the divided PCM signal. In addition, the band dividing filter 41 supplies the low pass PCM signal BS-L, which is a low pass component of the divided PCM signal, to the high pass component generator 42 and the band synthesis filter 43.

The high pass component generator 42 uses a low pass PCM signal (BS-L) supplied from the band division filter 41 and a high pass envelope (ENV-H) supplied from the inverse quantization unit 32, A PCM signal is generated to be a pseudo high pass PCM signal (BS-H). About the method of generating the pseudo high pass PCM signal (BS-H), it is described in patent document 1 which the applicant filed previously. The high pass component generator 42 supplies the pseudo high pass PCM signal BS-H to the band synthesis filter 43.

The band synthesis filter 43 synthesizes the low pass PCM signal BS-L supplied from the band division filter 41 and the pseudo high pass PCM signal BS-H supplied from the high pass component generator 42. The PCM signal of all bands is output as a decoding result.

As compared with the voice corresponding to the PCM signal without the high frequency component, the voice corresponding to the PCM signal of the entire band output as described above is reduced in feeling of frustration and becomes a clear and audible voice.

4 is a diagram for explaining signals output from the inverse MDCT unit 33 and the band synthesis filter 43. As shown in FIG. 4, the horizontal axis represents frequency and the vertical axis represents the level of the signal. This is the same also in FIG. 7, FIG. 10, and FIG. 12-FIG. 16 mentioned later.

The signal output from the inverse MDCT unit 33 is a PCM signal of low frequency spectrum (SP-L) denormalized using a low frequency envelope (ENV-L) as shown in A of FIG. In addition, the signal output from the band synthesis filter 43 uses a low pass spectrum (SP-L) PCM signal denormalized using a low pass envelope (ENV-L) as shown in B of FIG. 4 as a low pass component. It is a PCM signal which has a pseudo high pass PCM signal (BS-H) generated from a high pass envelope (ENV-H) and a low pass PCM signal (BS-L) as a high pass component.

FIG. 5 is a flowchart for describing decoding processing by the decoding device 30 of FIG. 3. This decoding process is started, for example, when the bit stream encoded by the encoding device 10 is input to the decoding device 30.

In step S31 of FIG. 5, the decomposition unit 31 converts the bit stream input to the decoding device 30 into a low pass envelope (ENV-L), a low pass spectrum (SP-L), and a high pass envelope (ENV-H). It decomposes | disassembles and supplies it to the dequantization part 32.

In step S32, the inverse quantization unit 32 performs inverse quantization with respect to each of the low pass envelope ENV-L, the low pass spectrum SP-L, and the high pass envelope ENV-H supplied from the decomposition unit 31. Do it. The dequantization unit 32 supplies the dequantized low pass envelope (ENV-L) and the low pass spectrum (SP-L) to the inverse MDCT unit 33, and supplies the high pass envelope ENV-H to the band extension unit 34. Supplies).

In step S33, the inverse MDCT unit 33 performs normalization on the low pass spectrum SP-L using the low pass envelope ENV-L supplied from the inverse quantization unit 32.

In step S34, the inverse MDCT unit 33 performs inverse MDCT on the low frequency spectrum SP-L which is a denormalized frequency domain signal to obtain a PCM signal that is a time domain signal. The inverse MDCT unit 33 supplies this PCM signal to the band extension unit 34.

In step S35, the band division filter 41 of the band extension part 34 divides the PCM signal supplied from the inverse MDCT part 33 into a high frequency component and a low frequency component. Then, the band split filter 41 discards the high frequency component of the divided PCM signal, and converts the low frequency PCM signal BS-L, which is the low frequency component of the divided PCM signal, into the high frequency component generator 42 and the band synthesis filter ( 43).

In step S36, the high pass component generation unit 42 supplies the low pass PCM signal BS-L supplied from the band division filter 41 and the high pass envelope ENV-H supplied from the inverse quantization unit 32. To generate a pseudo high pass PCM signal (BS-H). The high pass component generator 42 supplies the pseudo high pass PCM signal BS-H to the band synthesis filter 43.

In step S37, the band synthesis filter 43 is a low pass PCM signal BS-L supplied from the band division filter 41 and a pseudo high pass PCM signal BS-H supplied from the high pass component generator 42. ) Is synthesized to obtain a full-band PCM signal. The band synthesis filter 43 outputs the PCM signal of the entire band and finishes the processing.

The above-described band extension technology has already been used in the international high-efficiency advanced audio coding (HE-AAC) or LPEC (trademark) stereo high quality mode.

As described above, in the conventional band extension technique, the band extension process is performed as a post-process (post process) of the decoding process of the low-band spectrum (SP-L). As a result, the degree of freedom of the pseudo high pass PCM signal BS-H can be increased. That is, the pseudo high pass PCM signal BS-H can be generated from the low pass PCM signal BS-L which is a time domain signal rather than the low pass spectrum SP-L which is a frequency domain signal.

In addition, by freely setting the processing block size of the encoding process and the decoding process and the processing block size of the band extension process, the frequency analysis precision and the time decomposition precision can be optimized respectively.

In addition, when generating a pseudo high pass PCM signal by the method described in patent document 1, while generating a noise spectrum from a high pass envelope (ENV-H), a high pass envelope (ENV-H) and a low pass PCM signal ( A complex process of generating tonal spectra from BS-L) and comparing both spectra is required.

The processing for generating such a noise spectrum and a tone spectrum is an essential process for improving the matching accuracy of the low frequency spectrum and the high frequency spectrum, which are necessary for generating audio of high quality sound, and are described in Patent Documents 2 and 3 It is also performed in the decoding apparatus.

Japanese Patent No. 381770 Japanese Patent No. 3646938 Japanese Patent No. 3646939

As mentioned above, in the conventional band extension technology, research, development, and practical use are performed so that a band extension process may be performed as a post-process of the decoding process of low-band spectrum (SP-L). Therefore, the PCM signal of the full band is the band after the normal decoding processing by the decomposition part 31, the inverse quantization part 32, and the inverse MDCT part 33 ends (time T0 in the example of FIG. 3). It is output after the processing time by the expansion part 34 (in the example of FIG. 3, time T1).

This is not a big problem when the decoding device 30 is installed in a reproducing device that simply reproduces only audio. However, when the decoding device 30 is installed in a reproducing device that also reproduces a video in synchronism with audio, for example, the output time of the PCM signal of all bands is different in the case of performing only normal decoding and in case of performing band expansion. This makes it difficult to output video and audio synchronously.

In order to solve this problem, it is necessary to slow down the playback timing of the video. However, since the buffering of the video requires a large amount of memory compared to the audio, the resource is increased. It is also conceivable to shift the synchronization timing of the video and audio in advance, but it is difficult to always specify the optimum synchronization timing because it depends on the playback device whether only normal decoding or bandwidth expansion is performed.

In addition, the decoding device 30 needs to newly install the band extension unit 34 for band extension, so that resources increase as compared to a decoder that does not perform band extension.

As described above, in the decoding device performing band extension, it is required to reduce the delay time caused by the band extension and to suppress the increase of resources.

This invention is made | formed in view of such a situation, and it is possible to reduce the delay time by bandwidth expansion at the time of decoding, and to suppress the increase of the resource of a decoding side.

The decoding device according to the first aspect of the present invention is a low band envelope of a speech signal, a low band spectrum normalized using the low band envelope, a high band envelope of the voice signal, and a high frequency spectrum concentration. On the basis of acquisition means for acquiring as a coding result, generation means for generating a spectrum using the low-pass spectrum normalized among the coding results acquired by the acquiring means, the envelope of the high-pass, and the concentration level, The randomization means for randomizing the phase of the spectrum generated by the generating means, and the low-band envelope among the encoding results obtained by the obtaining means, denormalizes the low-band spectrum, and selects the random The spectrum randomized by the converting means or the spectrum generated by the generating means A decoder for combining the spectrum of the low-range normalization, and having a means for synthesizing the synthesized result to the spectrum of the entire band.

The decoding method and program of the first aspect of the present invention correspond to the decoding device of the first aspect of the present invention.

In the first aspect of the present invention, a low frequency envelope of a speech signal, a low frequency spectrum normalized using the low frequency envelope, a high frequency envelope of the speech signal, and a concentration spectrum of the high frequency spectrum of the speech signal are encoded results. The spectrum is generated using the low-band spectrum and the high-band envelope among the obtained encoding results, and the phase of the spectrum is randomized based on the concentration. Using the low-pass envelope, the low-spectrum spectrum is denormalized, the randomized spectrum or the generated spectrum and the de-normalized low-spectrum spectrum are synthesized, and the synthesis result is the spectrum of the entire band.

The decoding device according to the second aspect of the present invention includes acquisition means for acquiring a low-pass envelope of a speech signal, a low-pass spectrum normalized using the low-pass envelope, and a high-pass envelope of the speech signal as a coding result; Generation means for generating a spectrum using the low frequency spectrum normalized among the encoding results acquired by the acquiring means and the envelope of the high frequency, and the low frequency normalized among the encoding results acquired by the acquiring means. Determination means for determining a concentration degree of the low-band spectrum based on the spectrum, randomization means for randomizing the phase of the spectrum generated by the generation means based on the concentration degree determined by the determination means; By using the low-pass envelope of the encoding result obtained by the obtaining means, Denormalizes the base low-spectrum spectrum, synthesizes the spectrum randomized by the randomization means or the spectrum generated by the generation means, and the denormalized low-spectrum spectrum, and synthesizes the result of the full band. It is a decoding apparatus provided with the synthesis | combining means used as spectrum.

The decoding method and program of the second aspect of the present invention correspond to the decoding device of the second aspect of the present invention.

In the second aspect of the present invention, an envelope of a low pass of a speech signal, a spectrum of a low pass normalized using the envelope of the low pass, and an envelope of a high pass of the speech signal are obtained as encoding results, and among the obtained encoding results. Based on the normalized low-band spectrum and the high-band envelope, a spectrum is generated, and the concentration of the low-band spectrum is determined based on the normalized low-band spectrum among the obtained encoding results, and the determined concentration And based on the generated spectrum, the phase of the generated spectrum is randomized, the low-band spectrum is denormalized using the low-pass envelope of the obtained encoding result, and the randomized spectrum or the generated spectrum, The denormalized low frequency spectrum is synthesized, and the synthesis result is the specification of the whole band. It is a column.

The encoding device of the third aspect of the present invention includes determining means for determining a degree of concentration of the high frequency spectrum based on a high frequency spectrum of an audio signal, and an envelope of the low frequency spectrum and the high frequency range from the spectrum of the audio signal. Extraction means for extracting an envelope of the spectrum, normalization means for normalizing the low-band spectrum using the envelope of the low-spectrum spectrum, the concentration determined by the determining means, the low-spectrum extracted by the extraction means And a multiplexing means for multiplexing the envelope of the high frequency spectrum, the low frequency spectrum normalized by the normalizing means, and making an encoding result.

The encoding method and program of the third aspect of the present invention correspond to the encoding device of the third aspect of the present invention.

In the third aspect of the present invention, the concentration of the high frequency spectrum is determined based on the high frequency spectrum of the audio signal, and the envelope of the low frequency spectrum and the envelope of the high frequency spectrum are extracted from the spectrum of the audio signal. The low frequency spectrum is normalized using the envelope of the low frequency spectrum, the determined concentration, the envelope of the low frequency spectrum extracted and the envelope of the high frequency spectrum, and the normalized low frequency spectrum are multiplexed, This results in the encoding.

The decoding apparatus of the first or second aspect and the encoding apparatus of the third aspect may be independent apparatuses or may be internal blocks constituting one apparatus.

According to the first and second aspects of the present invention, it is possible to reduce the delay time due to the band extension during decoding and to suppress the increase of resources.

Further, according to the third aspect of the present invention, encoding can be performed so that the delay time due to the expansion of the band during decoding is reduced, so that an increase in resources on the decoding side is suppressed.

1 is a block diagram illustrating an example of a configuration of an encoding apparatus.
FIG. 2 is a flowchart for describing an encoding process by the encoding device of FIG. 1.
3 is a block diagram showing an example of the configuration of a decoding device.
4 is a diagram illustrating a signal output from an inverse MDCT unit and a band synthesis filter.
5 is a flowchart for describing a decoding process by the decoding device of FIG. 3.
6 is a block diagram showing a configuration example of a first embodiment of an encoding apparatus to which the present invention is applied.
FIG. 7 is a diagram illustrating signals output from the MDCT unit and the quantization unit of FIG. 6. FIG.
FIG. 8 is a flowchart for describing an encoding process by the encoding device of FIG. 6.
9 is a block diagram illustrating a configuration example of a decoding device that decodes a bit stream encoded by the encoding device of FIG. 6.
FIG. 10 is a view for explaining a signal output from the inverse MDCT unit in FIG. 9.
11 is a diagram illustrating a difference of decoding results with or without phase randomization.
It is a figure explaining the characteristic of the high frequency spectrum (SP-H).
It is a figure explaining the characteristic of the high frequency spectrum (SP-H).
It is a figure explaining the characteristic of the high frequency spectrum (SP-H).
It is a figure explaining the characteristic of the high frequency spectrum (SP-H).
It is a figure explaining the characteristic of the high frequency spectrum (SP-H).
FIG. 17 is a flowchart for describing a decoding process by the decoding device of FIG. 9.
It is a block diagram which shows the structural example of 2nd Embodiment of the decoding apparatus which applied this invention.
FIG. 19 is a flowchart for describing a decoding process by the decoding device of FIG. 18.
20 is a diagram illustrating a configuration example of a computer.

<1st embodiment>

[Configuration example of the first embodiment of the encoding apparatus]

6 is a block diagram showing a configuration example of a first embodiment of an encoding apparatus to which the present invention is applied.

The same code | symbol is attached | subjected to the structure similar to the structure of FIG. 1 among the structures shown in FIG. Overlapping descriptions are omitted as appropriate.

The configuration of the encoding device 50 of FIG. 6 differs from the configuration of FIG. 1 in that the quantization unit 51 and the multiplexer 52 are provided instead of the quantization unit 12 and the multiplexer 13. The encoding device 10 generates a bit stream by multiplexing a random flag RND (detailed later) in addition to the low pass envelope ENV-L, the low pass spectrum SP-L, and the high pass envelope ENV-H. do.

Specifically, the quantization unit 51 of the encoding device 50 is composed of a determination unit 61, an extraction unit 62, a normalization unit 63, and a partial quantization unit 64.

The determination part 61 is based on the high frequency spectrum SP-H of the spectrum SP supplied from the MDCT part 11, for example, according to following formula (1) of the high frequency spectrum SP-H Determine concentration D.

In formula (1), max (SP-H) represents the maximum value of the high frequency spectrum SP-H, and ave (SP-H) represents the average value of the high frequency spectrum SP-H.

According to Equation 1, when the tonality of the high frequency component of the audio to be encoded is high and there is a large bias in the distribution of the high frequency spectrum (SP-H), the concentration D becomes large and the noise of the high frequency component of the audio to be encoded is increased. When the height is high and the distribution of the high frequency spectrum SP-H is flat, the concentration D becomes small.

The determination unit 61 determines the random flag RND based on the concentration D. This random flag RND is a spectrum pseudo-random to the high-band spectrum SP-H generated from the low-band spectrum SP-L and the high-band envelope ENV-H during the band extension processing in the decoder described later. A flag indicating whether or not to randomize the phase.

For example, when the concentration degree D is larger than the threshold set in advance in the encoding device 50, that is, when the tone property of the high frequency spectrum SP-H is high, the random flag RND is not randomized. Is determined by 0. On the other hand, when the concentration degree D is equal to or less than a preset threshold value, that is, when the noise of the high frequency spectrum SP-H is high, the random flag RND is determined to be 1 indicating that it is randomized. The determination unit 61 supplies the determined random flag RND to the multiplexer 52.

The extraction section 62, like the quantization section 12 of FIG. 1, is enveloped from the high-band spectrum SP-H and the low-band spectrum SP-L, respectively, of the spectrum SP supplied from the MDCT section 11. Extract

The normalizer 63 normalizes the low-band spectrum SP-L using the low-pass envelope ENV-L, similarly to the quantizer 12.

The partial quantizer 64 quantizes the normalized low pass spectrum SP-L, and supplies the resulting low pass spectrum SP-L to the multiplexer 52. In addition, similar to the quantization unit 12, the partial quantization unit 64 quantizes the extracted high pass envelope ENV-H and low pass envelope ENV-L. The partial quantization unit 64 supplies the quantized high pass envelope ENV-H and the low pass envelope ENV-L to the multiplexer 52, similarly to the quantization unit 12.

The multiplexer 52 includes a random flag RND supplied from the determination unit 61 of the quantization unit 51, a low pass envelope ENV-L and a low pass spectrum SP- supplied from the partial quantization unit 64. L) and the high pass envelope (ENV-H) are multiplexed. The multiplexer 52 outputs the resulting bit stream. This bit stream is recorded in a recording medium (not shown) or transmitted to the decoding device.

[Description of Signal in Coding Device]

FIG. 7 is a diagram for explaining signals output from the MDCT unit 11 and the quantization unit 51 of the encoding apparatus 50 of FIG. 6.

As shown in A of FIG. 7, the spectrum SP output from the MDCT unit 11 is a spectrum of all bands. On the other hand, signals other than the random flag RND output from the quantization unit 51, as shown in B of FIG. 7, have a low-band spectrum (SP-L), a low-pass envelope (ENV-L), and a high-pass envelope ( ENV-H).

[Description of Processing of Encoding Apparatus]

FIG. 8 is a flowchart for describing an encoding process by the encoding device 50 of FIG. 6. This encoding process is started, for example, when a PCM signal of speech is input to the encoding apparatus 50. FIG.

In step S51 of FIG. 8, the MDCT part 11 performs MDCT with respect to the PCM signal which is the time-domain signal of the audio input to the encoding apparatus 50, similarly to the process of step S11 of FIG. Generate the phosphorus spectrum (SP). The MDCT unit 11 supplies the generated spectrum SP to the quantization unit 51.

In step S52, the determination part 61 of the quantization part 51 is based on the high frequency spectrum SP-H among the spectrums SP supplied from the MDCT part 11, and by the above formula (1), Determine the concentration D of the high spectrum (SP-H).

In step S53, the determination unit 61 determines the random flag RND based on the concentration degree D. The determination unit 61 supplies the determined random flag RND to the multiplexing unit 52, and the process proceeds to step S54.

Since the process of step S54-S56 is the same as the process of step S12-S14 of FIG. 2, description is abbreviate | omitted.

After the process of step S56, in step S57, the multiplexing unit 52 supplies a random flag RND, a low pass envelope ENV-L, a low pass spectrum SP-L, and a high pass envelope supplied from the quantization unit 51. Multiplex (ENV-H) and output the resulting bit stream. The process then ends.

[Configuration example of decoding device]

FIG. 9 is a block diagram illustrating a configuration example of a decoding device that decodes a bit stream encoded by the encoding device 50 of FIG. 6.

The decoding device 70 of FIG. 9 is comprised by the decomposition part 71, the inverse quantization part 72, the high frequency component generation part 73, the phase random part 74, and the inverse MDCT part 75. As shown in FIG. The decoding device 70 performs band extension processing simultaneously with the decoding processing of the low frequency spectrum SPL.

Specifically, the decomposition unit 71 (acquisition means) acquires a bit stream encoded by the encoding device 50 of FIG. 6. The decomposing unit 71 decomposes the bit stream into a random flag (RND), a low pass envelope (ENV-L), a low pass spectrum (SP-L), and a high pass envelope (ENV-H). To feed.

The inverse quantization unit 72, like the inverse quantization unit 32 of FIG. 3, has a low pass envelope (ENV-L), a low pass spectrum (SP-L), and a high pass envelope (ENV-H) supplied from the decomposition unit 71. Inverse quantization is performed for each of

The inverse quantization unit 72 supplies the inversely quantized low pass envelope (ENV-L) to the inverse MDCT unit 75, and supplies the low spectrum (SP-L) to the inverse MDCT unit 75 and the high frequency component generation unit 73. Supplies). In addition, the inverse quantization unit 72 supplies the high frequency envelope ENV-H to the high frequency component generation unit 73, and the inverse quantization unit 72 supplies the random flag RND to the phase random unit 74. Supply.

The high pass component generator 73 generates a high pass spectrum using the low pass spectrum SP-L and the high pass envelope ENV-H supplied from the inverse quantization unit 72 to form a pseudo high pass spectrum. Specifically, for example, the high frequency component generator 73 replicates the low frequency spectrum (SP-L), transforms the copied spectrum using the high frequency envelope (ENV-H), and sets it as a pseudo high frequency spectrum.

As a method of generating this pseudo high-band spectrum, for example, the method described in Patent Literature 1 filed by the present applicant may be used, or a method other than that may be used. The high pass component generator 73 supplies the generated pseudo high pass spectrum to the phase random unit 74.

The phase random part 74 randomizes the phase of the pseudo high frequency spectrum supplied from the high frequency component generation part 73 based on the random flag RND supplied from the inverse quantization part 72.

Specifically, when the phase random unit 74 is 1 indicating that the random flag RND is randomized, the phase random unit 74 randomizes the sign (+/-) of the pseudo high frequency spectrum by the following equation (2). do.

In Equation 2, SP-H represents a high frequency spectrum and i represents a spectrum number.

According to equation (2), the sign of the high-spectrum (SP-H) is randomly assigned to either -1 or 1 by multiplying "-1" by the number of lower 1 bits of the return value of the random function rand (). do.

On the other hand, when it is 0 indicating that the random flag RND is not randomized, the phase random unit 74 does not randomize the phase of the pseudo high pass spectrum.

The phase random part 74 supplies the pseudo high frequency spectrum whose phase was randomized or the pseudo high frequency spectrum which is not randomized to the inverse MDCT part 75.

The inverse MDCT portion 75 (synthesizing means) denormalizes the low frequency spectrum SP-L using the low frequency envelope ENV-L supplied from the inverse quantization portion 72. The inverse MDCT unit 75 synthesizes the denormalized low frequency spectrum (SP-L) and the pseudo high frequency spectrum supplied from the phase random unit 74. The inverse MDCT unit 75 performs inverse MDCT on the spectrum of the entire band which is the frequency domain signal obtained as a result of the synthesis, and obtains the PCM signal of the full band which is the time domain signal. The inverse MDCT unit 75 outputs the PCM signal of the entire band as a decoding result.

As described above, the decoding device 70 simultaneously generates a low frequency spectrum (SP-L) and generates a pseudo high frequency spectrum. Therefore, the time required for decoding in the decoding device 70 is approximately equal to the time required for decoding in the normal decoding device which performs only decoding. That is, in the decoding device 70 of FIG. 9, after the bit stream is input, the decoding result can be output after time T0. That is, in the decoding device 70, no delay due to band extension occurs.

[Explanation of Signal in Decoding Device]

FIG. 10 is a diagram for explaining a signal output from the inverse MDCT unit 75 of the decoding device 70 of FIG. 9.

The signal output from the inverse MDCT unit 75 includes the low-pass spectrum SP-L normalized using the low-pass envelope ENV-L as shown in FIG. 10, and the high-pass envelope shown in FIG. 10. PCV signal after frequency conversion of the result of synthesis of the pseudo high pass spectrum generated in the ENV-H) and the low pass spectrum (SP-L).

[Explanation of Effect by Randomizing Phase]

11-16 is a figure explaining the effect of the phase randomization by the phase random part 74 of FIG.

11 is a diagram illustrating a difference of decoding results with or without phase randomization.

As illustrated in FIG. 11, in the encoding device 50 of FIG. 6, the PCM signal is encoded for each section having a constant length called a frame. However, the frames are usually overlapped and set by 50%. Specifically, as shown in Fig. 11, the J-1st frame and the next Jth frame are set overlapping by 0.5 frames.

In FIG. 11, the case where the spectrum with high tonality is encoded as shown to the left of FIG. 11 is demonstrated.

In this case, as shown in the upper right corner of FIG. 11, if the phase of the spectrum is not randomized at the time of decoding the spectrum of the J-1st and Jth frames, the overlap period of the J-1st and Jth frames is determined. The phase of the spectrum is correctly restored by combining the spectra and codes of the J-1 &lt; th &gt; and J &lt; th &gt; frames. Therefore, the spectrum of the restored overlap period becomes a spectrum with high tone.

On the other hand, as shown in the lower right, if the phase of the spectrum is randomized at the time of decoding the spectrum of the J-1st and Jth frames, the signs of the spectra of the J-1st and Jth frames do not necessarily coincide. Do not. Therefore, the phase of the spectrum of the overlap period is not correctly restored. Therefore, the signal of the overlap period restored in the decoding device 70 becomes a spectrum in which the tone characteristics of the spectrum before encoding have collapsed.

When the tonality of the spectrum collapses, the energy that should be originally concentrated in the particular spectrum leaks into the surrounding spectrum. As a result, the peak (acid) of the spectrum is suppressed as compared with the original spectrum, and the energy that begins to leak to the surroundings pushes up the energy of the valley of the spectrum. As a result, the spectrum is noisy.

As described above, when the phase is randomized at the time of decoding, the spectrum having tone characteristics before encoding is converted into a spectrum having noise characteristics.

12-16 is a figure explaining the characteristic of the high frequency spectrum (SP-H).

As shown in FIG. 12A, when the tone of the low-band spectrum SP-L is high, the tone of the high-band spectrum SP-H is also high in many cases. This can be estimated from the fact that musical instruments such as wind instruments and string instruments emit sound waves combining a fundamental frequency and a harmonic component of an integer multiple thereof.

When the spectrum consisting of the high-tone low-band spectrum (SP-L) and the high-band spectrum (SP-H) is band extension coded, the pseudo high band spectrum simply returns the low band spectrum (SP-L) during band extension decoding. When generated by this, as shown in FIG. 12B, the pseudo high pass spectrum is a high tone spectrum. Therefore, the voice corresponding to the decoding result is an audio with a low sense of discomfort.

Therefore, the encoding device 50 of FIG. 6 sets the random flag RND to 0 when the concentration degree D is larger than a preset threshold value, that is, when the high frequency component of the audio to be encoded has tone. As a result, in the decoding device 70, since the phase of the pseudo high-band spectrum is not randomized, the voice corresponding to the decoding result is audibly low voice.

On the other hand, as shown in FIG. 13A and FIG. 14A, when the noise of the low pass spectrum (SP-L) is high, the higher the higher the noise, the higher the noise. In the case of cymbals or maracas, which emit noises such as hitting sounds or impact sounds having high noise, i.e., non-toning characteristics, the higher frequencies of vibrations propagate in the instruments, so the higher frequencies of the vibration element It can be estimated from the fact that the phases are entangled in complexity and the noise is increased.

When the spectrum consisting of the low-noise low-frequency spectrum SP-L and the high-frequency spectrum SP-H is thus subjected to band extension coding, as shown in B of FIG. 13, the low band spectrum SP- is used during band extension decoding. The pseudo high pass spectrum generated using L) becomes a spectrum with high noise. Therefore, whether the randomization of the pseudo high-band spectrum is performed as shown in FIG. 13B or the randomization is performed as shown in FIG. 14B, the noise of the pseudo high-band spectrum becomes high, and thus corresponds to the decoding result. The voice is audibly voiced with less discomfort.

However, even if the noise of a musical instrument, such as a cymbal or a maracas, is high, the low-frequency component may contain the tone vibration component. Moreover, the frequency of the sound of musical instruments, such as a cymbal and a maraca, is mainly high frequency, and the low frequency component may contain the other high tone voice. Therefore, as shown in FIG. 15A or FIG. 16A, even when the noise of the high frequency spectrum SP-H is high, the tone of the low frequency spectrum SP-L may be high.

When the spectrum consisting of the high-tone low-frequency spectrum (SP-L) and the high-noise high-frequency spectrum (SP-H) is band-extended coded, as shown in B of FIG. 15, the low-band spectrum at the time of band extension decoding The pseudo high pass spectrum generated using (SP-L) may contain tonal components. Therefore, if the phase of the pseudo high-band spectrum is not randomized as shown in Fig. 15B, the high-frequency voice corresponding to the decoding result has no inherent noise, and has tone like the low-band voice, It is audibly discordant voice.

On the other hand, when the phase of the pseudo high pass spectrum is randomized, even when the tonal component is included in the original pseudo high pass spectrum, the pseudo high pass spectrum after randomization has noise as shown in FIG. 16B. Therefore, the voice corresponding to the decoding result is an audio with a low sense of discomfort.

As described above, when the high-band spectrum SP-H has noise, when the low-band spectrum SP-L also has noise, it does not matter whether randomization is performed or not, but the low-band spectrum SP-L If)) has tone, it is necessary to perform randomization. Therefore, when the high-spectrum SP-H has noise, it is possible to always perform randomization, so that a decoding result with an aural sense of incongruity can be obtained on the basis of the concentration D.

Therefore, the encoding apparatus 50 of FIG. 6 sets the random flag RND to 1 when the concentration degree D is equal to or less than a preset threshold value, that is, when there is noise in the high frequency component of the audio to be encoded. As a result, in the decoding device 70, the phase of the pseudo high-band spectrum is randomized, so that the voice corresponding to the decoding result is audible with a low sense of discomfort.

In addition, since low-frequency, high-noise, high-tone voices are hardly present in the natural world, they are applied to a spectrum composed of a high-noise low-band spectrum (SP-L) and a high-tone high-frequency spectrum (SP-H). It does not take into account.

[Description of Processing of Decryption Apparatus]

FIG. 17 is a flowchart for describing decoding processing by the decoding device 70 of FIG. 9. This decoding process is started, for example, when the bit stream encoded by the encoding device 50 is input to the decoding device 70.

In step S71 of FIG. 17, the decomposition unit 71 obtains a bit stream encoded by the encoding device 50, and sets the bit stream as a random flag (RND), a low pass envelope (ENV-L), and a low pass spectrum. Dissolve into (SP-L) and high-pass envelopes (ENV-H). The decomposing unit 71 supplies the random flag RND, the low pass envelope ENV-L, the low pass spectrum SP-L, and the high pass envelope ENV-H to the inverse quantization unit 72.

In step S72, the inverse quantization unit 72 performs inverse quantization with respect to each of the low pass envelope ENV-L, low pass spectrum SP-L, and high pass envelope ENV-H supplied from the decomposition unit 71. Do it. The inverse quantization unit 72 supplies the inversely quantized low pass envelope (ENV-L) to the inverse MDCT unit 75, and supplies the low spectrum (SP-L) to the inverse MDCT unit 75 and the high frequency component generation unit 73. Supplies). In addition, the inverse quantization unit 72 supplies the high frequency envelope ENV-H to the high frequency component generation unit 73, and the inverse quantization unit 72 supplies the random flag RND to the phase random unit 74. Supply.

In step S73, the high pass component generator 73 generates a pseudo high pass spectrum using the low pass spectrum SP-L and the high pass envelope ENV-H supplied from the inverse quantization unit 72. The high pass component generator 73 supplies the generated pseudo high pass spectrum to the phase random unit 74.

In step S74, the phase random part 74 determines whether the random flag RND supplied from the inverse quantization part 72 is one. When it is determined in step S74 that the random flag RND is 1, in step S75, the phase random unit 74 of the pseudo high pass spectrum supplied from the high pass component generation unit 73 is expressed by the above expression (2). Randomize the phase. And the phase random part 74 supplies the pseudo high frequency spectrum whose phase was randomized to the inverse MDCT part 75, and advances a process to step S76.

On the other hand, when it is determined in step S74 that the random flag RND is not 1, that is, the random flag RND is 0, the phase random unit 74 does not randomize the phase of the pseudo high-band spectrum and inverts MDCT as it is. It supplies to the part 75. The process then proceeds to step S76.

In step S76, the inverse MDCT unit 75 denormalizes the low pass spectrum SP-L by using the low pass envelope ENV-L supplied from the inverse quantization unit 32.

In step S77, the inverse MDCT unit 75 synthesizes the pseudonormalized low frequency spectrum (SP-L) and the pseudo high frequency spectrum supplied from the phase random unit 74, and inverses the spectrum of the entire band obtained as a result. MDCT is performed to obtain PCM signals of all bands. The inverse MDCT unit 75 outputs the PCM signal of the entire band as a decoding result and ends the processing.

As described above, the decoding device 70 generates a pseudo high frequency spectrum by using the low frequency spectrum (SP-L) before the inverse MDCT, and adds the pseudo high frequency spectrum to the random flag RND determined based on the concentration of the high frequency spectrum SP-H. By randomizing the pseudo high-band spectrum, the high-band component of the spectrum of the audio to be encoded is restored.

Thereby, using the low frequency spectrum SP-L, a spectrum relatively consistent with the high frequency spectrum SP-H can be reconstructed as a high frequency component of the spectrum of the audio to be encoded. Therefore, by using the low-band spectrum (SP-L) to restore the high-band component of the spectrum of the audio to be encoded, the low-band spectrum (SP-L) decoding and band extension processing can be performed simultaneously, resulting in delay due to band extension. You can cut your time. As a result, the PCM signal of the audio of the full band which is not frustrated and is clear and audible is output as the decoding result after approximately the same time as that of the decoding apparatus which does not perform the band extension process.

In addition, since the decoding device 70 generates a pseudo high frequency spectrum with noise by randomizing the phase of the pseudo high frequency spectrum generated by using the low frequency spectrum (SP-L), the simple high frequency spectrum is simply a random high frequency spectrum. Compared with the case where the spectrum is generated as a spectrum, a pseudo high-band spectrum more consistent with the high-band spectrum (SP-H) can be generated.

In addition, since the decoding device 70 generates the low and high frequency components of the spectrum before the inverse MDCT, the band division filter 41 and the band synthesis filter (such as the decoding device 30 of FIG. 3) for the band extension processing. 43) need not be provided. Therefore, compared with the decoding apparatus 30 of FIG. 3, resources, such as a throughput, a circuit scale, and a code size, for band expansion processing can be reduced.

&Lt; Second Embodiment >

[Configuration example of second embodiment of decoding device]

It is a block diagram which shows the structural example of 2nd Embodiment of the decoding apparatus which applied this invention.

The same code | symbol is attached | subjected to the structure similar to the structure of FIG. 3 or FIG. 9 among the structure shown in FIG. Overlapping descriptions are omitted as appropriate.

The configuration of the decoding device 100 of FIG. 18 mainly includes the decomposition part 31 and the inverse quantization part 32 provided in place of the decomposition part 71 and the inverse quantization part 72 and the newly determined part 101. ) Is different from the configuration of the decoding device 70 in FIG. 9. The decoding device 100 determines the random flag RND based on the low-band spectrum SP-L included in the bit stream encoded by the encoding device 10 of FIG. 1.

Specifically, the determination unit 101 is based on the low-band spectrum SP-L dequantized by the inverse quantization unit 32, for example, by the following equation (3). Determine the concentration of D '.

In formula (3), max (SP-L) represents the maximum value of the low-pass spectrum SP-L, and ave (SP-L) represents the average value of the low-band spectrum SP-L.

According to Equation 3, when the tone of the low frequency component of the audio to be encoded is high and there is a large bias in the distribution of the low frequency spectrum (SP-L), the concentration D 'becomes large and the noise of the low frequency component of the audio to be encoded is increased. When this is high and the distribution of the low-pass spectrum SP-L is flat, the concentration degree D 'becomes small.

The determination unit 101 determines the random flag RND based on the concentration degree D '. Specifically, when the concentration degree D is larger than the threshold value preset in the decoding device 100, that is, when the tone property of the low-band spectrum SP-L is high, the determination unit 101 determines the random flag RND. Determine 0 as 0. On the other hand, when the concentration degree D 'is equal to or less than a preset threshold value, that is, when the noise of the low-band spectrum SP-L is high, the determination unit 101 determines the random flag RND as one. The determination unit 101 then supplies the determined random flag RND to the phase random unit 74. As a result, the phase of the pseudo high frequency spectrum is not randomized when the tone of the low frequency spectrum SP-L is high, and the phase of the pseudo high frequency spectrum is randomized when the noise of the low frequency spectrum SP-L is high. . As a result, the voice corresponding to the decoding result is an audio of sufficient audio quality.

[Description of Processing of Decryption Apparatus]

FIG. 19 is a flowchart for describing a decoding process by the decoding device 100 of FIG. 18. This decoding process is started when the bit stream encoded by the encoding apparatus 10 of FIG. 1 is input into the decoding apparatus 100, for example.

In step S91 of FIG. 19, the decomposition unit 31 performs the low-pass envelope (ENV-L), low-band spectrum (SP-L), and high-pass envelope (ENV-H) on the bit stream encoded by the encoding device 10. Is decomposed to and supplied to the inverse quantization unit 32.

Since the process of step S92 and S93 is the same as the process of step S72 and S73 of FIG. 17, description is abbreviate | omitted.

After the process of step S93, in step S94, the determination part 101 is based on the low-pass spectrum SP-L dequantized by the inverse quantization part 32, and according to the above formula (3), the low-pass spectrum Determine the concentration D 'of (SP-L).

In step S95, the determination unit 101 determines the random flag RND based on the concentration degree D '. The determination unit 101 supplies the random flag RND to the phase random unit 74, and advances the process to step S96.

Since the process of step S96-S99 is the same as the process of step S74-S77 of FIG. 17, description is abbreviate | omitted.

&Lt; Third Embodiment >

[Description of Computer to which the Present Invention]

Subsequently, the above-described series of encoding processing and decoding processing may be performed in hardware or in software. When a series of encoding and decoding processes are performed by software, the program constituting the software is installed in a general-purpose computer or the like.

Therefore, FIG. 20 shows the structural example of one Embodiment of the computer in which the program which performs a series of process mentioned above is installed.

The program can be recorded in advance in the storage unit 208 or ROM (Read Only Memory) 202 serving as a recording medium built in a computer.

Alternatively, the program can be stored (recorded) in the removable media 211. Such removable media 211 can be provided as so-called package software. Here, the removable media 211 includes, for example, a flexible disk, a compact disc read only memory (CD-ROM), a magneto optical (MO) disk, a digital versatile disc (DVD), a magnetic disk, a semiconductor memory, and the like.

In addition to installing the above-described removable media 211 to a computer via the drive 210, the program can be downloaded to a computer via a communication network or a broadcasting network and installed in the built-in storage unit 208. Can be. That is, the program can be wirelessly transmitted to the computer through a satellite for digital satellite broadcasting, for example, from a download site, or wired to the computer via a network such as a local area network (LAN) or the Internet.

The computer has a CPU (Central Processing Unit) 201 built in, and the CPU 201 is connected to the input / output interface 205 via the bus 204.

The CPU 201 executes a program stored in the ROM 202 when a command is input through the input / output interface 205 by the user by operating the input unit 206 or the like. Alternatively, the CPU 201 loads and executes a program stored in the storage unit 208 into the RAM (Random Access Memory) 203.

As a result, the CPU 201 performs the processing according to the above-described flowchart or the configuration of the above-described block diagram. The CPU 201 transmits the processing result to the output unit 207 or from the communication unit 209 via the input / output interface 205 as necessary, for example, to the storage unit 208. Record it.

In addition, the input unit 206 includes a keyboard, a mouse, a microphone, or the like. The output unit 207 is configured of an LCD (Liquid Crystal Display), a speaker, or the like.

Here, in the present specification, the processing performed by the computer according to the program does not necessarily need to be performed in time series in the order described as the flowchart. That is, the processing performed by the computer according to the program also includes processing executed in parallel or separately (for example, parallel processing or processing by an object).

The program may be processed by one computer (processor) or may be distributedly processed by a plurality of computers. In addition, the program may be transmitted to a remote computer and executed.

Embodiment of this invention is not limited to embodiment mentioned above, A various change is possible in the range which does not deviate from the summary of this invention.

50: encoding device
52: multiplexer
61: decision
62: extraction unit
63: normalization unit
70: decoding device
71: decomposition
73: high frequency component generation unit
74: phase random portion
75: reverse MDCT department
100: decoding device
101: decomposition unit
101: decision

Claims (14)

As a decoding apparatus,
Acquiring means for acquiring a low band envelope of the speech signal, a low band spectrum normalized using the low band envelope, a high band envelope of the voice signal, and a concentration of the high band spectrum of the voice signal as encoding results;
Generation means for generating a spectrum using the normalized low frequency spectrum and the high frequency envelope among the encoding results obtained by the acquisition means;
Randomization means for randomizing the phase of the spectrum generated by the generation means based on the concentration degree;
The low-frequency spectrum is denormalized by using the low-pass envelope among the encoding results obtained by the acquiring means, and the spectrum generated by the randomizing means or the spectrum generated by the generating means; And synthesizing means for synthesizing the denormalized low frequency spectrum and making the synthesis result the spectrum of the entire band. 2. The method according to claim 1, wherein the randomization means does not randomize the phase of the spectrum generated by the generating means when the concentration is greater than a predetermined threshold and the concentration is less than or equal to the predetermined threshold. And a decoding device for randomizing the phase of the spectrum generated by the generating means. The method of claim 1,
The acquiring means acquires a random flag that is information indicating whether the randomization means determined based on the low envelope, the low spectrum, the high envelope, and the concentration degree to randomize,
When the randomization means is information indicating that the random flag is randomized, the phase of the spectrum is randomized and supplied to the synthesizing means, and when the information indicates that the random flag is not randomized, the spectrum The decoding device which supplies to the said synthesis | combining means, without randomizing the phase of a. As a decoding method,
The decoding apparatus comprises:
An acquisition step of acquiring, as an encoding result, a low pass envelope of the speech signal, a low pass spectrum normalized using the low pass envelope, an envelope of the high pass of the speech signal and a concentration of the high pass of the speech signal as an encoding result;
A generating step of generating a spectrum using the normalized low frequency spectrum and the high frequency envelope among the encoding results obtained by the obtaining step;
A randomization step of randomizing a phase of the spectrum generated by the processing of the generating step based on the concentration level;
The low-band spectrum is denormalized by using the low-pass envelope of the encoding result obtained by the processing of the acquiring step, and the process of the spectrum or the generation step randomized by the processing of the randomization step. And a synthesis step of synthesizing the spectrum generated by the low frequency spectrum and the normalized low frequency spectrum, and making the synthesis result the spectrum of the entire band. On the computer,
An acquisition step of acquiring, as an encoding result, a low pass envelope of the speech signal, a low pass spectrum normalized using the low pass envelope, an envelope of the high pass of the speech signal and a concentration of the high pass of the speech signal as an encoding result;
A generating step of generating a spectrum using the normalized low frequency spectrum and the high frequency envelope among the encoding results obtained by the obtaining step;
A randomization step of randomizing a phase of the spectrum generated by the processing of the generating step based on the concentration level;
The low-band spectrum is denormalized by using the low-pass envelope of the encoding result obtained by the processing of the acquiring step, and the process of the spectrum or the generation step randomized by the processing of the randomization step. And a synthesizing step of synthesizing the spectrum generated by the denormalized low-pass spectrum and making the synthesis result the spectrum of the entire band. As a decoding apparatus,
Acquisition means for acquiring a low-pass envelope of the speech signal, a low-pass spectrum normalized using the low-pass envelope and a high-pass envelope of the speech signal as a coding result;
Generation means for generating a spectrum using the normalized low frequency spectrum and the high frequency envelope among the encoding results obtained by the acquisition means;
Determination means for determining a degree of concentration of the low frequency spectrum based on the normalized low frequency spectrum among the encoding results obtained by the obtaining means;
Randomization means for randomizing a phase of the spectrum generated by the generation means based on the degree of concentration determined by the determination means;
The low-frequency spectrum is denormalized by using the low-pass envelope among the encoding results obtained by the acquiring means, and the spectrum generated by the randomizing means or the spectrum generated by the generating means; And synthesizing means for synthesizing the denormalized low frequency spectrum and making the synthesis result the spectrum of the entire band. 7. The method according to claim 6, wherein the randomization means does not randomize the phase of the spectrum generated by the generation means when the concentration is greater than a predetermined threshold and the concentration is less than or equal to the predetermined threshold. And a decoding device for randomizing the phase of the spectrum generated by the generating means. The method according to claim 6,
The determining means further indicates that the randomization means does not randomize a random flag, which is information indicating whether the randomization means randomizes when the concentration of the low frequency spectrum is greater than a predetermined threshold. Information is determined, and when the concentration of the low frequency spectrum is equal to or less than the predetermined threshold value, the random flag is determined as information indicating that the randomization means is randomized,
When the randomization means is information indicating that the random flag is randomized, the phase of the spectrum is randomized and supplied to the synthesizing means, and when the information indicates that the random flag is not randomized, the spectrum The decoding device which supplies to the said synthesis | combining means, without randomizing the phase of a. As a decoding method,
The decoding apparatus comprises:
An acquisition step of acquiring, as an encoding result, a low pass envelope of the speech signal, a low pass spectrum normalized using the low pass envelope, and an envelope of the high pass of the speech signal as encoding results;
A generating step of generating a spectrum using the normalized low frequency spectrum and the high frequency envelope among the encoding results obtained by the obtaining step;
A determining step of determining a degree of concentration of the low frequency spectrum based on the low frequency spectrum normalized among the encoding results obtained by the processing of the obtaining step;
A randomization step of randomizing a phase of the spectrum generated by the processing of the generating step based on the concentration degree determined by the processing of the determining step;
The low-band spectrum is denormalized by using the low-pass envelope of the encoding result obtained by the processing of the acquiring step, and the process of the spectrum or the generation step randomized by the processing of the randomization step. And a synthesis step of synthesizing the spectrum generated by the low frequency spectrum and the normalized low frequency spectrum, and making the synthesis result the spectrum of the entire band. On the computer,
An acquisition step of acquiring, as an encoding result, a low pass envelope of the speech signal, a low pass spectrum normalized using the low pass envelope, and an envelope of the high pass of the speech signal as encoding results;
A generating step of generating a spectrum using the normalized low frequency spectrum and the high frequency envelope among the encoding results obtained by the obtaining step;
A determining step of determining a degree of concentration of the low frequency spectrum based on the low frequency spectrum normalized among the encoding results obtained by the processing of the obtaining step;
A randomization step of randomizing a phase of the spectrum generated by the processing of the generating step based on the concentration degree determined by the processing of the determining step;
The low-band spectrum is denormalized by using the low-pass envelope of the encoding result obtained by the processing of the acquiring step, and the process of the spectrum or the generation step randomized by the processing of the randomization step. And a synthesizing step of synthesizing the spectrum generated by the denormalized low-pass spectrum and making the synthesis result the spectrum of the entire band. As an encoding device,
Determination means for determining a degree of concentration of the high frequency spectrum based on the high frequency spectrum of the audio signal;
Extraction means for extracting an envelope of a low frequency spectrum and an envelope of the high frequency spectrum from the spectrum of the speech signal;
Normalization means for normalizing the low frequency spectrum using an envelope of the low frequency spectrum;
And encoding the concentration degree determined by the determining means, the envelope of the low frequency spectrum extracted by the extraction means and the envelope of the high frequency spectrum, and the low frequency spectrum normalized by the normalization means to produce an encoding result. An encoding device comprising multiplexing means. The method of claim 11,
The concentration determining means further determines whether or not to randomize the spectrum when the decoding device for decoding the encoding result generates a predetermined spectrum as the high frequency spectrum when the concentration is greater than a predetermined threshold. Determine the random flag as the information to indicate that the random flag is not randomized; when the concentration is less than or equal to the predetermined threshold value, determine as the information indicating to randomize the random flag;
And the multiplexing means multiplexes the random flag, the envelope of the low-spectrum spectrum, the envelope of the high-spectrum spectrum, and the normalized low-band spectrum to form the encoding result. As a coding method,
An encoding apparatus comprising:
A determining step of determining the concentration of the high frequency spectrum based on the high frequency spectrum of the speech signal,
An extraction step of extracting an envelope of the low frequency spectrum and an envelope of the high frequency spectrum from the spectrum of the speech signal;
A normalization step of normalizing the low frequency spectrum using an envelope of the low frequency spectrum,
Multiplexing the concentration determined by the processing of the determining step, the envelope of the low frequency spectrum and the envelope of the high frequency spectrum extracted by the processing of the extraction step, and the spectrum of the low frequency normalized by the processing of the normalization step And a multiplexing step of making an encoding result. On the computer,
A determining step of determining the concentration of the high frequency spectrum based on the high frequency spectrum of the speech signal,
An extraction step of extracting an envelope of the low frequency spectrum and an envelope of the high frequency spectrum from the spectrum of the speech signal;
A normalization step of normalizing the low frequency spectrum using an envelope of the low frequency spectrum,
Multiplexing the concentration determined by the processing of the determining step, the envelope of the low frequency spectrum and the envelope of the high frequency spectrum extracted by the processing of the extraction step, and the spectrum of the low frequency normalized by the processing of the normalization step To execute a process including a multiplexing step of making an encoding result.

Images